Added support for Homebrew MacOSX packaging system and for Eigen3 (in Eigen2 compatibility mode)

The cause for this is actually a bug in Boost.Python that triggers when
arg("arg")... is used in the export and the number of arg("") arguments
doesn't match the number of argument of the function. When one then uses
a wrong kwarg, the program crashes. I'll submit a bug report to the
Boost.Python on this issue. I fear that there are many more cases in
the pymod export where we are using arg("") as we did for the query.

This is based on exhaustive testing on the complete files of the PDB.
There are still a few unresolved issues. However it is not clear to me,
whether the problems are due to errors in the PDB files or our
misunderstanding.

The residue names listed in the SEQRES record get automatically translated to
one-letter-codes. Requires the presence of the rule-based builder to work
properly. Turned off by default

Thanks Andi!

move ClashScore to mol.alg, AminoAcid to conop

Previously we were returning the wrong index for residues
in chains that contained a certain residue number more
than once. We were always returning the index of the first
residue. Among other things, this fixes BZDNG-229.